Transformation of response variable = subtract pre-test score from outcome score which generates a “gain score” = g(i) = alpha + T(zi) + error(i). As indicated in the chapter, the gain scores essentially assume that Beta = 1 in the model. If one were to include the pre-test score as a predictor -> g(i) = alpha + TZ(i) + yx(i) + error(i) - then the estimate for the the coefficient for z is equivalent to the estimated coefficient from the original model, y(i) = alpha + TZ(i) + betax(i) + error(i).``
Exercise 4 & 5, Chapter 18